In the past few days two studies on the impact of music streaming on recorded music sales surfaced. The Country Music Association (CMA) presented study results that streaming services are more successful in fostering music sales than radio. It was reported that more than a quarter of the respondents purchased music after streaming it compared to eight percent of radio listeners. Since the CMA study is not available, the results cannot be seriously assessed. Therefore, we rely on a study of Stephen McBride, a researcher of the Science Team at Pandora Media. He testified before the U.S. Copyright Royalty Judges and presented evidence that show a positive impact on music sales. In a nutshell, he highlights that listening to a song on Pandora increases music sales be more than 2 percent – a moderate “Pandora effect”, as he called it. In following, I will analyse McBride’s findings in the so-called Music Sales Experiments that were conducted by Pandora’s science team. Learn more about it here:
Pandora’s “Music Sales Experiments”
Pandora conducts controlled experiments to test what effect the use of the music streaming service has on recorded music sales. The research team, therefore, manipulate the availability of certain songs on Pandora in different regions in the U.S. This allows to compare the effects on music sales in each region and answers the question if there is a difference or not. The research group conducted several experiments from June to September 2014. Within 8 weeks new titles as well as catalog titles in the Pandora repertoire were tested. It was disabled to stream these titles for 8 weeks in a randomly selected U.S. region. 814 experiments were made with tracks that were new to Pandora and 401 experiments involved catalog titles. The New Music Sales Experiments (NMSEs) included only album tracks that were new to Pandora during the selection week as well as older versions of the same track with at least one sale in the observation period. Tracks from Greatest Hits albums and other compilations were excluded from the experiments. In the Catalog Music Sales Experiments (CMSEs) 200 songs were randomly selected from the Rolling Stone Top 500 list that were played more than 1,000 times in June 2014 and had more than 25 sales in the previous week. In a second round again 200 titles from the Pitchfork 500 Songs were selected with more than 700 plays in the week prior to the experiment and with more than 25 sales in the previous week. Then SoundScan data, which are available for 230 U.S. sales regions, were used to assess the impact of the experiments on the tracks’ unit sales (physical and digital).
The 841 New Music Sales Experiments show that plays of an album track on Pandora increase its sales by an average of 2.31 percent. The positive sales’ effect in the Catalog Music Sales Experiments is even higher with plus 2.66 percent. Both results are statistically significant. The research team also compared tracks from albums/songs released from major music companies with those from indies. A positive promotional effect of 2.82 percent was measured for new music from major record companies. For catalog music from the majors the effect is still positive with an increase of 2.36 percent. However, the New Music Sales Experiments for tracks released by indie labels showed no significant effect, whereas Pandora increases the sales of catalog songs by indie labels by 3.85 percent. Therefore, McBride concludes: “The results show that, even for music that is broadly familiar, playing on Pandora increases sales of music” (McBride 2014: 19).
Table 1: Overview of the Music Sales Experimentations results
In addition, the research team also tried to assess a revenue effect for labels and artists of a song’s stream (spin) on Pandora. The experiments show that a stream of a major label’s song on Pandora generates additional 0.167 cents of sales revenue for the label. Plays of a catalog recording by a major company results in an increase of 0.006 cents per stream and an increase of 0.008 cents for an indie company respectively.
McBride eventually concluded: “The Music Sales Experiments confirm that the Pandora radio is promotional of music sales – that is, music sales are higher when music plays on Pandora. (…) We also present evidence that the promotional effect is greater for music with greater exposure on Pandora.” (McBride 2014: 20).
It is obvious that the study is instrumental before the US Copyright Royalty Court to lobby for lowering the licensing fees Pandora has to pay to the copyright holders. Nevertheless, the experimental setup is useful to measure the impact of music streaming on recorded music sales. Since Pandora is a mainly free service, the study supports the results of file sharing studies that did not find any negative impact of file sharing on music sales. The study shows a weak promotional effect of free music streaming on music sales and should be taken into consideration when the freemium models are taken into question. It would interesting if studies on Spotify, Deezer and other streaming service would come to same results.
Billboard.biz, “Streaming Drives Sales, According to New Study”, November 20, 2014, http://www.billboard.com/articles/business/6327135/streaming-drives-sales-according-to-new-study (accessed November 27, 2014).
McBride, Stephan, 2014, Written direct testimony in the matter of determination of rates and terms for digital performance in sound recordings and ephemeral recordings before the United States Copyright Royalty Judges, October 7, 2014.